Catch-Phrase: Truthful Text-to-SQL. Instantly.

Catch-Phrase is a front-end for natural language phrasal search, analytics, and update over SQL-based databases. Catch-Phrase is launched directly in your AWS account, giving you full control over configuration, data and security. Unlike LLM-based methods, Catch-Phrase uses combinatorial search to find all the meaningful interpretations of the user's prompt. If there are multiple possible interpretations or if the user is requesting a data modification, then Catch-Phrase paraphrases interpretations back to the user in English for selection or confirmation. This gives users full control over what SQL gets applied. Unlike LLM-based methods, Catch-Phrase answers the bulk of user questions in under one second. And it never hallucinates.

Non-technical users can 'google' the database using precise, often complex conditions. Results are specific values, lists, tables or charts. Users can share their questions in a single click and export results to CSV files for use in other tools. Shared questions become regular HTML links that can be embedded in documents and webpages so that when clicked upon, they let others execute the shared question over the database. Often users simply modify and click on canned common questions, integrity checks, KPI reports or data entry or modification forms. Users can also often be led through special data stories that take them through a predefined sequence of natural language prompts. Even the most non-technical users can find this point and click set-up very easy to use.

Catch-Phrase supports flexible cloud-based administration. Just log into the Admin Interface, connect to your remote database and automatically build an initial interface in under five minutes. Use editing tools to customize and refine an interface with synonyms, regular expressions, ellipsis, additional views and reports. Set up some corpora of common questions, integrity checks, KPI reports or data entry or modification forms. Launch a natural language interface and invite your users to access it. Now wait and explore and manage XML-based logs of user sessions to pinpoint where to improve your interface. Relaunch and repeat.

Each new release of Catch-Phrase comes with example configurations and databases. Launching Catch-Phrase will give you immediate access to these examples which you are free to adapt to your own purposes. The number of users and natural language interfaces per virtual machine is limited only by the size of the virtual machine you launch. It is quite likely all your needs will be met with a single small virtual machine.

At a technical level Catch-Phrase covers a wide fragment of SQL. Specifically it supports SQL select, insert, update and delete queries: Catch-Phrase can map natural language to OLTP and OLAP queries expressing regular conditions, set conditions, date and time conditions, superlatives, negations, universal and existential quantification, cardinality conditions, etc. Catch-Phrase also integrates well with transaction processing and database check-pointing via explicit SQL logging and the ability to start, abort or commit transactions.

Catch-Phrase supports open formats and APIs for full interoperability: All configuration for Catch-Phrase natural language interfaces is represented in a single XML file in a completely documented and open format. In addition, live natural language interfaces can be accessed via REST calls. This lets you embed natural language interfaces across many applications including those that use voice recognition. Finally, Catch-Phrase can easily be secured with certificates to allow for HTTPS access.